Ajay Arora

Also super interested in this... feels like most generations are way too slow speaking wise and sound nothing like a natural conversation. Any immediate solutions I can try?

i have a feeling your learning rates might be too high? thought it would be 1e-4 or so.

@yawnzh did you ever figure out a workaround for this? want to train musicgen on cloud GPUs that don't have SLURM set up.

running into the same issue